github-starred/cs249r_book
2
0
Fork 0
mirror of https://github.com/harvard-edge/cs249r_book.git synced 2026-05-21 13:31:55 -05:00
Code Issues 114 Packages Projects Releases 24 Wiki Activity

[PR #1782] fix(labs): dark-mode selector, lab11 MathPeek, vol2-lab06 tau_opt #15733

New Issue
Open
opened 2026-05-20 14:04:55 -05:00 by GiteaMirror · 0 comments
GiteaMirror commented 2026-05-20 14:04:55 -05:00
Owner
Copy Link

📋 Pull Request Information

Original PR: https://github.com/harvard-edge/cs249r_book/pull/1782
Author: @Shashank-Tripathi-07
Created: 5/18/2026
Status: 🔄 Open

Base: dev ← Head: fix/labs-audit-bugs

📝 Commits (3)

  • fbb3f3d fix(site): add nav-footer and dropdown-menu dark mode selectors
  • f265b2e fix(labs): replace @media dark-mode selector in 404.qmd and fix lab11 MathPeek
  • 97dbe71 fix(labs): correct tau_opt values in vol2 lab06 Part B checkpoint storm

📊 Changes

5 files changed (+57 additions, -25 deletions)

View changed files

📝 labs/404.qmd (+10 -12)
📝 labs/vol1/lab_11_hw_accel.py (+1 -1)
📝 labs/vol2/lab_06_fault_tolerance.py (+8 -8)
📝 shared/config/footer-site.yml (+1 -1)
📝 shared/styles/_site-dark.scss (+37 -3)

📄 Description

Summary

Three verifiable content/rendering bugs found during audit of all labs (vol1 + vol2).

  • 404.qmd dark-mode: @media (prefers-color-scheme: dark) is ignored by Quarto; should be body.quarto-dark. Ten CSS rules converted. Same pattern as existing fixes in other site components.
  • vol1/lab11 MathPeek (Part E, FlashAttention tiling): Text said "Speedup ≈ M/d, which is 2-4x at typical dimensions" -- stale from before the Part E answer key was corrected from option C to option D in #1676. At the default settings (seq=4096, tile=256) the simulation computes ~21x capped to 10x, making option D (~10x) correct. Updated explanation to use the correct formula (Speedup ≈ N/B) and value (~10x).
  • vol2/lab06 Part B (checkpoint storm): Concept framing gave MTBF as ~5 hours for a 1,000-GPU cluster but GPU_MTTF_HOURS = 50,000 makes that 50 hours (5 hours is the 10,000-GPU case). The correct-answer callout then showed two wrong tau_opt values:
    • 1,000-GPU case: claimed ~2.6 h, actual = sqrt(2 × 2450 × 180,000) ≈ 8.2 h
    • 10,000-GPU case: claimed ~27 min, actual = sqrt(2 × 2450 × 18,000) ≈ 156 min (~2.6 h)
      Both the framing and the callout are corrected.

Bugs skipped (already merged to dev)

fix/lab01-esp32-sram-oom-ratio, fix/lab02-esp32-sram-field, fix/lab03-esp32-sram-field, fix/lab04-cascade-amp-factor, fix/lab05-flop-scaling-answer, fix/lab07-esp32-sram-field, fix/lab11-part-e-tiling-answer, fix/lab13-cold-start-answer, fix/lab14-debt-multiplier-answer

Test plan

  • Build labs site; verify 404 page renders dark-mode styles in dark theme
  • Open lab11 Part E with seq=4096, tile=256; confirm MathPeek now says ~10x
  • Open vol2 lab06 Part B with default sliders (175B model, 1,000 GPUs, 1 GB/s NFS); confirm displayed tau_opt ≈ 8.2 h, callout text matches

🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.

## 📋 Pull Request Information **Original PR:** https://github.com/harvard-edge/cs249r_book/pull/1782 **Author:** [@Shashank-Tripathi-07](https://github.com/Shashank-Tripathi-07) **Created:** 5/18/2026 **Status:** 🔄 Open **Base:** `dev` ← **Head:** `fix/labs-audit-bugs` --- ### 📝 Commits (3) - [`fbb3f3d`](https://github.com/harvard-edge/cs249r_book/commit/fbb3f3d1638fd7b123d885c9cbcd1cbd7e8257cc) fix(site): add nav-footer and dropdown-menu dark mode selectors - [`f265b2e`](https://github.com/harvard-edge/cs249r_book/commit/f265b2e994402cde50527e4d4613794e0ec4c768) fix(labs): replace @media dark-mode selector in 404.qmd and fix lab11 MathPeek - [`97dbe71`](https://github.com/harvard-edge/cs249r_book/commit/97dbe714c23a1acbb50b488754d5152fc523d566) fix(labs): correct tau_opt values in vol2 lab06 Part B checkpoint storm ### 📊 Changes **5 files changed** (+57 additions, -25 deletions) <details> <summary>View changed files</summary> 📝 `labs/404.qmd` (+10 -12) 📝 `labs/vol1/lab_11_hw_accel.py` (+1 -1) 📝 `labs/vol2/lab_06_fault_tolerance.py` (+8 -8) 📝 `shared/config/footer-site.yml` (+1 -1) 📝 `shared/styles/_site-dark.scss` (+37 -3) </details> ### 📄 Description ## Summary Three verifiable content/rendering bugs found during audit of all labs (vol1 + vol2). - **404.qmd dark-mode**: `@media (prefers-color-scheme: dark)` is ignored by Quarto; should be `body.quarto-dark`. Ten CSS rules converted. Same pattern as existing fixes in other site components. - **vol1/lab11 MathPeek (Part E, FlashAttention tiling)**: Text said _"Speedup ≈ M/d, which is 2-4x at typical dimensions"_ -- stale from before the Part E answer key was corrected from option C to option D in #1676. At the default settings (seq=4096, tile=256) the simulation computes ~21x capped to 10x, making option D (~10x) correct. Updated explanation to use the correct formula (Speedup ≈ N/B) and value (~10x). - **vol2/lab06 Part B (checkpoint storm)**: Concept framing gave MTBF as ~5 hours for a 1,000-GPU cluster but `GPU_MTTF_HOURS = 50,000` makes that 50 hours (5 hours is the 10,000-GPU case). The correct-answer callout then showed two wrong tau_opt values: - 1,000-GPU case: claimed ~2.6 h, actual = sqrt(2 × 2450 × 180,000) ≈ 8.2 h - 10,000-GPU case: claimed ~27 min, actual = sqrt(2 × 2450 × 18,000) ≈ 156 min (~2.6 h) Both the framing and the callout are corrected. ## Bugs skipped (already merged to dev) fix/lab01-esp32-sram-oom-ratio, fix/lab02-esp32-sram-field, fix/lab03-esp32-sram-field, fix/lab04-cascade-amp-factor, fix/lab05-flop-scaling-answer, fix/lab07-esp32-sram-field, fix/lab11-part-e-tiling-answer, fix/lab13-cold-start-answer, fix/lab14-debt-multiplier-answer ## Test plan - [ ] Build labs site; verify 404 page renders dark-mode styles in dark theme - [ ] Open lab11 Part E with seq=4096, tile=256; confirm MathPeek now says ~10x - [ ] Open vol2 lab06 Part B with default sliders (175B model, 1,000 GPUs, 1 GB/s NFS); confirm displayed tau_opt ≈ 8.2 h, callout text matches --- <sub>🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.</sub>
GiteaMirror added the pull-request label 2026-05-20 14:04:55 -05:00
Sign in to join this conversation.
No Branch/Tag Specified
Branches Tags
Labels
Clear labels
area: book
area: collabs
area: kits
area: labs
area: socratiq
area: tinytorch
area: tools
area: website
bug
dependencies
format: epub
format: pdf
javascript
link-health
link-rot
priority-high
pull-request
Mirrored from GitHub Pull Request
 staffml
type: bug
type: citation
type: code
type: errata
type: improvement
type: new
type: question
vault-sli
No Label pull-request
Milestone
No items
No Milestone
Projects
Clear projects
No project
Assignees
Clear assignees
GiteaMirror ninjasurge
No Assignees
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: github-starred/cs249r_book#15733
Delete Branch "%!s()"

Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?